The short version
Two of the best-known vulnerability research teams now assume attackers can turn a patch into an exploit almost immediately, so they will no longer keep your bugs private for three or four months. If your company ships software, a report from Google or Anthropic soon comes with a 30- to 60-day public clock. Teams that run penetration testing and security audits on a yearly cycle should plan for fixes measured in weeks.
What did Google Project Zero change?
Project Zero dropped its fixed patch cushion and will publish bug details a set number of days after it reports them, whether or not a fix has shipped.
Since 2021 the team gave vendors 90 days to fix a bug plus up to 30 more days for users to install the patch. The new policy, published on November 13, 2026 by researcher Natalie Silvanovich, calls this model “Full X”: disclosure X days after the report, with X falling in steps.
- From November 13, 2026: Full 90.
- From February 2, 2027: Full 60.
- From January 3, 2028: Full 30.
Most special cases are gone. Only three exceptions remain:
- bugs already exploited in the wild;
- bugs others have already made public;
- cases where the vendor asks for earlier disclosure.
Project Zero says it talked to major vendors first; BleepingComputer reports that some warned an instant switch to 30 days could expose unfixed bugs while they rebuild their processes, which is why the cut is phased.
What did Anthropic change?
Anthropic will publish high and critical bugs its models find in other companies’ software and in open source projects 30 days after reporting them instead of 90, and all other bugs after 45 days.
The new terms, announced on November 16, apply to reports sent on or after January 1, 2027. Anthropic also tightened two follow-up steps: if a maintainer does not answer, it will escalate to an outside coordinator after 15 days instead of 30, and once a patch exists it will wait 30 days instead of 45 before publishing full technical details.
The company explained the move in one line: “These policies assumed that fixes moved at human speed and that independent rediscovery of the same bug was unlikely. Neither assumption holds today.” Hard-to-patch software such as critical infrastructure, and flaws that need a specification or RFC change, can still get custom timelines.
How do the new disclosure deadlines compare?
| Rule | Project Zero (old) | Project Zero (new) | Anthropic (old) | Anthropic (from Jan 1, 2027) |
|---|---|---|---|---|
| Standard deadline | 90 days + up to 30 for patch rollout | 90 → 60 (Feb 2027) → 30 days (Jan 2028), fixed or not | 90 days | 30 days high/critical, 45 days other |
| Exploited in the wild | 7 days | 7 days | 7 days | 7 days |
| No response from maintainer | n/a | n/a | Escalate after 30 days | Escalate after 15 days |
| Full details after a patch | Within the 30-day cushion | At the deadline | 45 days | 30 days |
Why are researchers moving faster now?
Both teams point to AI. Project Zero writes that large language models have increased the speed at which attackers find and exploit vulnerabilities, and that treating patch-to-exploit time as zero is now a reasonable model of attacker behavior. It also says it now finds several bugs a week in targets where it once found a few a year.
Fresh evidence arrived the same week. On November 16, Google’s Threat Intelligence Group described an attacker framework called exploit_agent that compared web application code before and after security patches, located the fixed flaw and ran the exploit it generated without a human in the loop. A long private window protects nobody if a model can rebuild the bug from the patch diff.
What it means for US & EU software teams
The practical change is that the public clock now starts when a bug is reported, not when you finish fixing it.
If your product receives bug reports
- Triage in days. A 30-day deadline leaves little room for a two-week wait before an engineer looks at the report.
- Ship fixes outside the release train. Monthly or quarterly releases do not fit a 30-day window. You need a path to push a security fix to production, app stores and on-prem customers within a sprint.
- Publish a working security contact. Anthropic now escalates to an outside coordinator after 15 days of silence, so an unmonitored inbox costs real time.
If you depend on open source and vendors
Expect more bugs to become public before every downstream product has patched them. Dependency updates, container rebuilds and redeploys need to be routine, automated work rather than a project. In the EU this lines up with the Cyber Resilience Act, whose duty to report actively exploited vulnerabilities applies from September 11, 2026.
What should engineering teams do before January?
- Measure your real fix time. Take the last five security bugs and count days from report to production. Anything over 30 is the gap to close.
- Set severity SLAs. Agree in writing how fast high and critical issues get an owner, a fix and a release.
- Automate patch delivery. Make dependency bumps, rebuilds and rollouts a pipeline job with tests, not a manual task.
- Check your disclosure page. Publish a security.txt file and a monitored contact so reports reach engineers on day one.
- Plan for hard-to-patch parts. Firmware, mobile apps behind store review and customer-hosted installs need their own fast path.
Frequently asked questions
What is Project Zero’s new disclosure policy?
Google Project Zero replaced its “90 + 30” policy on November 13, 2026 with fixed deadlines. It now publishes bug details 90 days after reporting them to the vendor, whether or not a fix has shipped. The deadline drops to 60 days on February 2, 2027 and to 30 days on January 3, 2028.
What did Anthropic change in its vulnerability disclosure policy?
For reports sent on or after January 1, 2027, Anthropic will publicly disclose high and critical vulnerabilities its models find after 30 days instead of 90, and all other bugs after 45 days. It will escalate to an outside coordinator after 15 days without a response and publish full technical details 30 days after a patch.
Do the new deadlines apply to bugs that are already being exploited?
Actively exploited bugs keep the existing 7-day window at both Project Zero and Anthropic. The shorter standard deadlines apply to bugs that are not yet known to be exploited.
Why are disclosure deadlines getting shorter?
Both organisations say AI models now find and exploit vulnerabilities much faster, and that a bug one model finds is likely to be found by another. Project Zero says treating patch-to-exploit time as zero is now a reasonable model of attacker behavior.
What should a software vendor do to prepare?
Measure how long security fixes currently take from report to production, set written SLAs for high and critical issues, automate dependency updates and rollouts, and publish a monitored security contact so reports reach engineers on the first day.
Sources
Google Project Zero: Vulnerability Disclosure Policy, 2026 Edition (primary, November 13, 2026)
Anthropic: Coordinated vulnerability disclosure timelines (primary, November 16, 2026)
BleepingComputer: Google Project Zero cuts bug disclosures to 30 days amid AI threats (November 2026)
Help Net Security: Anthropic vulnerability disclosure deadline drops to 30 days for serious bugs (November 17, 2026)
SecurityWeek: Google Project Zero sets course for 30-day disclosure deadline (November 2026)
Google Threat Intelligence Group: agentic threat activity report (November 16, 2026)